PHYSICAL REVIEW B VOLUME 57, NUMBER 1 1 JANUARY 1998-I Current-induced resistivity switching effects near the melting line in detwinned YBa2Cu3O72d S. N. Gordeev,* D. Bracanovic, A. P. Rassau, and P. A. J. de Groot Department of Physics, University of Southampton, Southampton, SO17 1BJ, United Kingdom R. Gagnon and L. Taillefer Department of Physics, McGill University, Montreal (Quebec), Canada H3A 2T8 ~Received 22 September 1997! Hysteresis of the E-J characteristics has been observed in the vicinity of the melting line using a sensitive superconducting quantum interference device picovoltmeter with millikelvin temperature resolution. It was found that in this region the system could be switched from a lower- to a higher-resistivity state by applying a current in excess of a threshold value. From the comparison of transport and ac susceptibility data, we conclude that in this hysteretic region the vortex solid and liquid phases coexist and that the observed transport phenomena can be explained in terms of the rearrangement of solidified vortex domains. @S0163-1829~98!07201-4# Recent calorimetric1 and magnetization2 studies have unambiguously shown that, in clean YBa2Cu3O72d crystals, the phase transition from the vortex liquid to the vortex solid state is first order. In transport measurements this transition is seen as a sharp drop in the tail of both the temperature and field dependences of the resistivity together with associated hysteresis.3–7 Safar et al.3 have attributed this hysteresis to the supercooling and superheating effects which normally accompany a first-order phase transition. Charalambous et al.6 found that changes in the probing current only had an effect on the r (T) curves obtained on heating and not those obtained on cooling, thereby concluding that the vortex system can be superheated, but not supercooled. More recently, computer simulations by Dominguez et al.8 have indicated that a sufficiently large transport current could induce melting of the vortex lattice. To investigate the effect of the probing currents close to the melting line, we have performed a detailed study of the E-J curves in addition to more conventional r (T) measurements. We explicitly demonstrate that the E-J curves in this region show hysteretic ‘‘switching’’ behavior. From a comparative study of r (T) and x 8 (T), we conclude that the vortices in the hysteretic region are not in a single state, rather that in this region, vortex liquid and solid phases coexist. We have also observed that a current density, in excess of some threshold value, can switch the vortex system from the branch of the resistivity hysteresis curve obtained on heating to that obtained on cooling. We interpret this observation in terms of current-induced melting of superheated regions of the vortex solid. Our experiments were performed on an YBa2Cu3O72d single crystal ~of dimensions 1.1 mm30.35 mm365 mm!, grown in an Y-stabilized ZrO2 crucible and detwinned as described elsewhere.9 A superconducting quantum interference device ~SQUID! picovoltmeter was used to perform magnetoresistance measurements using a conventional fourprobe technique ~with H i c!. The probing current (J i a) was modulated with a single-polarity square wave of frequency f m 568 Hz. Resistivity versus temperature hysteresis loops were obtained by warming and cooling the sample at a con- stant rate of 0.01 K/min. During E-J measurements, the temperature stability was ;1 mK. ac susceptibility measurements were performed using a coaxial mutual inductance system with superimposed ac and dc fields.10 For J50.44 A/cm2 the sample under investigation exhibited Ohmic r (T) behavior very similar to that observed in previous experiments.3,4 The resistivity below the melting line dropped to zero over very narrow temperature interval (,0.03 K), and the r (T) dependence was found to be hysteretic in this vicinity ~see Fig. 1!. Notice in particular that, as is usually observed,3–7 the hysteresis loop was asymmetric: The curve obtained on warming was smoother than that obtained on cooling. For current densities less than a threshold value J t 50.6 A/cm2, the shape of the r (T) curves and the width of the hysteresis were independent of the current density. Figure 2 presents a set of isothermal E-J curves for B 52 T. The ‘‘cooled-state’’ curves ~lines! were obtained by cooling from the vortex liquid state with subsequent curves separated by a temperature step DT55 mK. Each separate curve was obtained by sweeping the current density from 0 to 10 A/cm2. The symbols represent the E-J curves for a 0163-1829/98/57~1!/645~4!/$15.00 645 57 FIG. 1. Temperature dependence of the normalized resistivity at B52 T, showing the hysteresis around T m . © 1998 The American Physical Society 646 S. N. GORDEEV et al. FIG. 2. E-J characteristics in the vicinity of T m . The lines represent the E-J curves of the cooled state for temperatures in the range 88.955–89.015 K separated by 5-mK steps. The symbols show the E-J curves for the heated state at T588.995 K for four successive runs. Each separate curve was obtained by sweeping the current density from zero up to a progressively higher maximum value J max : ~d! 0.5 A/cm2, ~3! 0.75 A/cm2, ~s! 1.0 A/cm2, and ~1! 1.5 A/cm2. ‘‘heated state’’ obtained by warming the sample from 88 K ~where the vortex system is in the solid state! up to 88.995 K. This state was then investigated by probing the E-J dependences in the regime of increasing J for four successive runs such that each subsequent run reached a higher maximum current density J max . It can be seen from Fig. 2 that provided J did not exceed a threshold value J t ('0.6 A/cm2), the heated state was stable with a reproducible E-J dependence located well below the cooled-state curve for the same temperature (T588.995 K). The shift between these states corresponds to a temperature difference ;8 mK, which is in agreement with the width of the r (T) hysteresis ~as shown in Fig. 1!. However, if the current during a particular run exceeded J t , then subsequent runs demonstrated a higher Ohmic resistivity r l 5E/J at low currents. The higher the probing current density, the larger was the r l value. For J.2J t saturation occurred, after which the E-J curve reproduced the shape of the cooled-state curve at the same temperature, but with a small vertical shift. This shift can be explained in terms of the temperature reproducibility of our system, which is ;2 mK for 1-K temperature cycles. It is therefore clear that a sufficiently large current density can switch the vortex system from a low-resistivity heated state to a highresistivity cooled state.11 As shown in Fig. 3, similar ‘‘switching’’ behavior was observed when the heated state was prepared by sweeping B at constant T. The solid lines show the E-J curves for the cooled state at B53 T. A heated state was obtained by sweeping the magnetic field from 3.0 to 2.5 T and then back to 3 T ~with T587.33 K!. This state was probed by increasing the current density J from 0 to 10 A/cm2 ~solid circles!. A subsequent current sweep ~open circles! coincided closely with the cooled-state curve for the same temperature. This demonstrates once more the switching effect displayed in Fig. 2. A cooled state was prepared by increasing B up to 3.5 57 FIG. 3. Demonstrates the effect of probing current on the E-J curves for B53 T. The solid lines show cooled-state E-J curves for temperatures in the range 87.29–87.36 K separated by 5-mK steps. The circles show two successive runs for the heated state at T 587.330 K @~d! first run, ~s! second run#. The crosses show two successive runs for the cooled state at T587.335 K @~1! first run, ~3! second run#. T and dropping back to 3 T ~with T587.335 K!. The stability of this state is demonstrated by the similarity of the E-J curves for two successive current sweeps ~vertical and diagonal crosses!. Jiang et al.7 have proposed the following explanation for the r (T) hysteresis observed in the vicinity of the melting line. On cooling the vortex system, just below T m a finite shear modulus develops, but initially this is insufficient to impede the motion of the vortices. The system will therefore supercool until the shear elastic energy exceeds the work done by the transport current. In contrast, when the sample is warmed up starting from the vortex solid state, initially the vortex velocities are lower. This means that less work is done by the transport current Lorentz force, and thus the vortex lattice should remain intact to higher temperatures. Our experiments demonstrate that the resistivity hysteresis shown in Fig. 1 is not due to the mechanism proposed by Jiang et al.7 Measurements were performed using a singlepolarity square-wave current, and thus for half of each modulation period the current flowing through the sample was zero. During these half-periods, the work done by the current was also zero. It follows that at temperatures for which the shear modulus is finite the vortex system should freeze. Thus, for this particular case, there should be no difference between the transition temperature obtained on heating and that obtained on cooling according to the mechanism proposed by Jiang et al. Figure 4 compares the resistivity r (T) and the real part of the ac susceptibility x 8 (T) in the vicinity of the melting line ~at B52 T!. From this it can be seen that the vortex-liquid– to–solid transition manifests itself as a simultaneous sharp change in both r (T) and x 8 (T) dependences. Notice that in contrast to x 8 (T) measurements performed on less clean samples ~which do not show a first-order melting transition!, we have observed a well-defined linear drop in x 8 (T). 10 The temperature width of the linear region (DT tr50.35 K) was independent of the applied ac field and therefore on the magnitude of the induced currents. These results are strikingly 57 CURRENT-INDUCED RESISTIVITY SWITCHING . . . FIG. 4. Comparison of the temperature dependence of the resistivity ~for J50.44 A/cm2! and ac susceptibility ~for ac fields of 50 and 25 mT! at B52 T. All of the curves were obtained in the regime of increasing temperature. The dashed lines delimit the transition region where the liquid and solid phases coexist. The dotted lines drawn along the ac susceptibility curves in the transition region are a guide to the eye. similar to those recently obtained by Fendrich et al.12 In a comparative study of transport and magnetic properties, they found that the onset of the resistivity and magnetization transitions coincided closely and that the reversible dc magnetization showed a sharp linear increase over a temperature interval ;0.2 K. They associated this rise with an increase in the fraction of the solid phase. Thus transport phenomena within this region can be understood in terms of coexistent vortex solid and liquid phases. Furthermore, they deduced from their results that the Ohmic resistivity dropped to zero where the fraction of solid phase reached 20% and explained this in terms of percolative processes.13 While ac susceptibility measurements do not reflect equilibrium ~thermodynamic! quantities, there is compelling evidence to suggest that the linear sections of the x 8 (T) dependences can be identified with the temperature region over which the liquid and solid vortex phases coexist. Of great significance is the fact that the linear region of the x 8 (T) dependences shows a remarkable correspondence, both in shape and position, to the transition region as observed by Fendrich et al. in dc magnetization measurements. Moreover, the ac susceptibility transitions are very well defined with onsets which coincide closely with the start of the resistivity transition for the same sample. The S-shaped E-J curves shown in Fig. 2 can also be explained in terms of coexistent vortex liquid and solid phases. In the low-current region of the curves, the response is Ohmic, suggesting that at this limit the dissipative contribution due to thermally activated motion of the vortex solid is negligible.14 Thus the implication is that the low-current Ohmic part of the E-J curves is controlled by vortex liquid flowing in channels between static solid domains. As T decreases, these domains grow and as a result the Ohmic resistivity is observed to fall. This resistivity finally drops to zero upon closing of the last channel between domains.15 At high currents the E-J curves are again Ohmic, but with a much higher resistivity r ff due to Bardeen-Stephen free flux flow of practically all the vortices.5 In the region of 647 intermediate current densities, the E-J curves show strong upward curvature. As seen from Fig. 2, the transition from Ohmic to non-Ohmic behavior at low currents is not a smooth crossover as would be the case for a transition from thermally assisted flux flow ~TAFF! to a flux creep regime. Instead, the nonlinear regime starts at a well-defined crossover current density J cr ~0.7 A/cm2 for T589.995 K!. We suggest that this crossover marks the point at which the vortices within the solid domains start to move. The central result of our paper is as follows. We have found that the threshold density J t , at which current begins to affect the ‘‘heated state,’’ almost coincides with J cr at which the E-J curve ~for the same temperature! deviates from Ohmic behavior ~see Fig. 2!. This implies that both effects are consequences of the same dynamic process. Thus the threshold current for the ‘‘switching’’ effect also marks the point at which the vortex-solid domains start to move. Our understanding of the current-induced switching effect is based on the assumption that the local melting temperature varies throughout the sample. This is supported by our ac susceptibility data,10 which showed that the width of the melting transition region (DT tr), where the vortex liquid and solid phases coexist, varied between 0.35 and 0.55 K depending on the particular sample. This implies that factors such as sample inhomogeneity lead to a variation of the local melting temperature within a given sample. Thus, as the melting line is approached from the direction of the vortex solid, domains of solidified vortices will persist at points where the local melting temperature T Lm is highest. The eventual result will be a number of isolated domains surrounded by channels of free flux flow. It is expected that in the heated state there will be regions within solid or possibly whole domains which are superheated relative to the temperature of the region in which they are located.3–5 We propose that under the influence of transport current Lorentz forces ~for J.J cr! the solid domains will be displaced into regions where T Lm is lower, leading to partial or complete melting. Resolidification is expected to take place at the peaks in the melting temperature profile, but not at points which were originally occupied by superheated solid. Thus, on considering the properties of the low-current Ohmic state, the net result will be a reduction in the fraction of vortices in the solid phase. This accounts for the order of magnitude difference between the dissipation in the original low-current state and the state obtained upon switching ~see Fig. 2!. It is also clear from this explanation why a transport current causes switching from the heated state to the cooled state, but not vice versa: A transport current can induce melting, but not solidification. Recent numerical simulations by Dominguez et al.8 have suggested that sufficiently large currents, much larger than the depinning current, should induce melting due to the blowing out of thermally induced vortex loops. It is, however, clear from our experiments that there is an additional distinct mechanism of current-induced melting. We have seen that this mechanism becomes effective at much lower currents. Using the model of coexistent phases, we can now explain the observed r (T) behavior in the vicinity of T m . For J ,J cr the current is insufficient to move the solid domains, resulting in current-independent resistivity hysteresis. For J .J cr a switching effect occurs and the resistivity depen- 648 S. N. GORDEEV et al. dences of the heating and cooling cycles collapse into a single r (T) curve. In addition, our interpretation also explains the lack of time dependence observed by Jiang et al.7 in their r (T) measurements. In the hysteretic region energy dissipation is largely due to the flow of vortex liquid between solid domains. Thus the system should not evolve with time ~as would be expected for a ‘‘glassy’’ solid phase16!. In conclusion we have demonstrated, from the analysis of E-J curves in the r (T) hysteretic region, that a sufficiently large probing current density ~J.1.5 A/cm2 for B52 T! can switch the vortex system from a lower-resistivity state obtained on heating into a higher-resistivity state. This higherresistivity state was shown to be the same as the state obtained on cooling from the vortex liquid state. The switching *On leave from the Moscow Institute of Radioengineering, Electronics and Automation, 117454 Moscow, Russia. 1 A. Schilling, R. A. Fisher, N. E. Phillips, U. Welp, D. Dasgupta, W. K. Kwok, and G. W. Crabtree, Nature ~London! 382, 791 ~1996!; M. Roulin, A. Junod, A. Erb, and E. Walker, J. Low Temp. Phys. 105, 1099 ~1995!. 2 T. Nishizaki, Y. Onodera, and N. Kobayashi, Phys. Rev. B 53, 82 ~1996!; R. Liang, D. A. Bonn, and W. N. Hardy, Phys. Rev. Lett. 76, 835 ~1996!; U. Welp, J. A. Fendrich, W. K. Kwok, G. W. Crabtree, and B. W. Veal, ibid. 76, 4809 ~1996!. 3 H. Safar, P. L. Gammel, D. A. Huse, D. J. Bishop, W. C. Lee, J. Giapintzakis, and D. M. Ginsberg, Phys. Rev. Lett. 69, 824 ~1992!; 70, 3800 ~1993!. 4 W. K. Kwok, S. Flesher, U. Welp, V. M. Vinokur, J. Downey, G. W. Crabtree, and M. M. Miller, Phys. Rev. Lett. 69, 3370 ~1992!. 5 W. K. Kwok, J. Fendrich, U. Welp, S. Fleshler, J. Downey, and G. W. Crabtree, Phys. Rev. Lett. 72, 1088 ~1994!; 72, 1092 ~1994!. 6 M. Charalambous, J. Chassy, P. Lejay, and V. Vinokur, Phys. Rev. Lett. 71, 436 ~1993!. 7 W. Jiang, N-C. Yeh, D. S. Reed, U. Kriplani, and F. Holtzberg, Phys. Rev. Lett. 74, 1438 ~1995!. 8 D. Dominguez, N. Gro” nbech-Jensen, and A. R. Bishop, Phys. Rev. Lett. 75, 4670 ~1995!; 78, 2644 ~1997!. 57 effect was explained within the framework of a model of coexistent phases, as supported by our ac susceptibility results. According to this explanation, the probing current moved solidified vortices to regions where the local melting temperature was lower, thereby inducing melting of superheated sections of the vortex solid. This project is supported by the EPSRC ~UK!. We appreciate helpful discussions with M. Oussena and are grateful to V. B. Geshkenbein, A. A. Zhukov, M. Charalambous, and M. V. Indenbom for useful comments. D.B. acknowledges funding from the University of Southampton. R.G. and L.T. acknowledge funding from NSERC ~Canada!, FCAR of Quebec, and CIAR ~Canada!. L.T. acknowledges the support of the Sloan Foundation. 9 R. Gagnon, C. Lupien, and L. Taillefer, Phys. Rev. B 50, 3458 ~1994!. 10 D. Bracanovic, S. N. Gordeev, S. Pinfold, R. Langan, M. Oussena, and P. A. J. de Groot, Physica C ~to be published!. 11 In principle, such effects could be induced by Joule heating in the sample contacts. However, since Joule heating is proportional to J 2 , the resultant temperature increase should lead to a pronounced bending of all the E-J curves, contrary to what we have observed. 12 J. Fendrich, U. Welp, W. K. Kwok, A. E. Koshelev, G. W. Crabtree, and B. W. Veal, Phys. Rev. Lett. 77, 2073 ~1996!. 13 G. Deutscher, in Percolation, Localisation and Superconductivity, Vol. B109 of NATO Advanced Study Institute, Series B: Physics, edited by A. M. Goldman and S. A. Wolf ~Plenum, New York, 1984!, p. 94. 14 If there were a significant contribution to the dissipation due to the motion of the vortex solid, then we would expect the response to be highly nonohmic. This can be clearly seen from the response at higher currents for which there is definitely dissipation due to the motion of both vortex solid and liquid. 15 In the mixed-state regime, the flow of current is expected to be highly nonuniform, and thus within this region the current density J reflects the average value across the whole cross section of the sample rather than the value in any particular region. 16 V. B. Geshkenbein, L. B. Ioffe, and A. I. Larkin, Phys. Rev. B 48, 9917 ~1993!.